Iannucci, David. The Hachijo Language of Japan: Phonology and Historical Development. 2019. hh:phonv:Iannucci:Hachijo
s > t / _s is probably better stated as Qs > ts.
The words _uQcei_ 'day before yesterday' (SJ _ototoi_, Hachijo < *ututui?), _ceitachi_ 'first of the month' (SJ _tsuitachi_), and _cei_ 'staff, stick' (SJ _tsue_) are analyzed as containing the syllable /tu/ (> [cu] as in SJ). _cei_ forms a minimal pair with _tei_ 'one day' and _kei_ 'today'. However, Iannucci says _cei_ "seems clearly to correspond to an underlying [cui] or [cue], based on Japanese cognates". This is unclear: do these sequences ever appear on the phonetic level? If so, it seems better to analyze /ts/ as phonemic, even though it only otherwise appears as an allophone of /s/ after sokuon or an allophone of /t/ before /u/.